---
license: mit
task_categories: [tabular-classification, time-series-forecasting]
tags: [econophysics, market-efficiency, randomness, honest-odds]
pretty_name: NoEdge Structure Dataset
---

# NoEdge Structure Dataset

The labeled data behind the **[No-Edge Detector](https://huggingface.co/nexusfinancial-dev/no-edge-detector)**.
Each row is the causal-feature fingerprint of a 256-point price window, labelled by its true generating
process: **memoryless_no_edge** (random walk), **mean_reverting** (Ornstein–Uhlenbeck), or **trending**
(AR(1) momentum). 9000 rows, balanced.

Features: acf1, acf2, acf3, acf4, acf5, absacf1, absacf2, absacf3, vr2, vr4, vr8, hurst, runs_z, er (autocorrelation of returns & |returns|, variance ratios, Hurst,
runs-test z, efficiency ratio).

Part of **Nexus — The Honest Odds Project**. It captures *statistical structure, not profit* — structure
does not imply a tradeable edge after payout. Not financial advice. Reproduce with `build_dataset.py`.